Thrips are one of the major worldwide crop pest that can be found in wide range of crops. Thrips are small in size (~1mm) and differences in appearance between some key species can be difficult to discern. However, their accurate identification to species level is essential to agriculture. Hence, this repository contains a software that cable of classifying two major thrip pest species - Western Flower Thrips (WFTs) and Plague Thrip. The software built upone two main modules - a Data Processing module and a Domain Knowledge-Driven Stacked Model. The Data Preprocessing Module segments relevant insect features and splits the insect into body segments to inform identification. The Domain Knowledge-Driven Stacked Model generates the prediction from each body segment and fuses predictions for each segment into an accurate species-level classification.
In addition, we provide with you a dataset that consists of microscopic images of the two thrip species to train and test the models in https://drive.google.com/drive/folders/1vfWZVaIwxsLgQG6CzE_8zLDMJfCBtA6S?usp=sharing.
The pretrained model can be found in https://drive.google.com/drive/folders/1hQR7v0s5gdwIuLM84T-cYpaBePUQqVVx?usp=sharing
- Open the WesternFlowerThrip_or_PlagueThrips file using Jupyter Notebook or Colab Notebook
- Replace the PATH with the path to the image folder you want to test
- Replace the path_for_models with the path to the pre-trained models
- Direct path_to_save to a temporary folder where you would like to save temporary results
- Run the code